U.S. patent application number 17/529516 was filed with the patent office on 2022-03-10 for spoofing detection device, spoofing detection method, and recording medium.
This patent application is currently assigned to NEC Corporation. The applicant listed for this patent is NEC Corporation. Invention is credited to Yusuke MORISHITA.
Application Number | 20220075999 17/529516 |
Document ID | / |
Family ID | |
Filed Date | 2022-03-10 |
United States Patent
Application |
20220075999 |
Kind Code |
A1 |
MORISHITA; Yusuke |
March 10, 2022 |
SPOOFING DETECTION DEVICE, SPOOFING DETECTION METHOD, AND RECORDING
MEDIUM
Abstract
To provide a technology of more accurately detecting spoofing in
face authentication, without increasing a scale of a device
configuration and a burden on a user. A spoofing detection device
includes a facial image sequence acquisition unit, a line-of-sight
change detection unit, a presentation information display unit, and
a spoofing determination unit. The facial image sequence
acquisition unit acquires a facial image sequence indicating the
face of a user. The line-of-sight change detection unit detects
information about a temporal change in the line-of-sight from the
facial image sequence. The presentation information display unit
displays presentation information presented to the user as part of
an authentication process. The spoofing determination unit
determines the likelihood of the face indicated by the facial image
sequence being spoofing on the basis of the information about the
temporal change in the line-of-sight with respect to the
presentation information.
Inventors: |
MORISHITA; Yusuke; (Tokyo,
JP) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
NEC Corporation |
Tokyo |
|
JP |
|
|
Assignee: |
NEC Corporation
Tokyo
JP
|
Appl. No.: |
17/529516 |
Filed: |
November 18, 2021 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
16682083 |
Nov 13, 2019 |
11210500 |
|
|
17529516 |
|
|
|
|
15514573 |
Mar 27, 2017 |
10515262 |
|
|
PCT/JP2015/005147 |
Oct 9, 2015 |
|
|
|
16682083 |
|
|
|
|
International
Class: |
G06K 9/00 20060101
G06K009/00; G06T 7/00 20060101 G06T007/00 |
Foreign Application Data
Date |
Code |
Application Number |
Oct 15, 2014 |
JP |
2014-210529 |
Claims
1. A face detection device comprising: at least one memory
configured to store instructions; and at least one processor
configured to execute the instructions to perform: acquiring a
facial image sequence indicating a user's face; displaying
presentation information on a display device and changing a
position of the presentation information on a screen of the display
device; detecting, from the facial image sequence, a direction of
line-of-sight of the user; and determining a face of the user
indicated by the facial image sequence is a living face, in
accordance with relationship between temporal change of the
direction of the line-of-sight of the user and the change of the
position of the presentation information on the screen of the
display device.
2. The face detection device according to claim 1, wherein the
presentation information includes textual information.
3. The face detection device according to claim 1, wherein the
presentation information includes a diagram.
4. The face detection device according to claim 1, wherein the at
least one processor is configured to execute the instructions to
perform: determining a display timing and a display position of the
presentation information at random.
5. The face detection device according to claim 1, wherein the at
least one processor is configured to execute the instructions to
perform: determining the face of the user is a living face, in
accordance with relationship between (i) the temporal change of the
direction of the line-of-sight of the user and a temporal change of
position coordinates of a point on a screen of the display device
closely watched by the user, and (ii) the change of the position of
the presentation information on the screen of the display
device.
6. A face detection method comprising: acquiring a facial image
sequence indicating a user's face; displaying presentation
information on a display device and changing a position of the
presentation information on a screen of the display device;
detecting, from the facial image sequence, a direction of
line-of-sight of the user; and determining a face of the user
indicated by the facial image sequence is a living face, in
accordance with relationship between temporal change of the
direction of the line-of-sight of the user and the change of the
position of the presentation information on the screen of the
display device.
7. The face detection method according to claim 6, wherein the
presentation information includes textual information.
8. The face detection method according to claim 6, wherein the
presentation information includes a diagram.
9. The face detection method according to claim 6, comprising:
determining a display timing and a display position of the
presentation information at random.
10. The face detection method according to claim 6, comprising:
determining the face of the user is a living face, in accordance
with relationship between (i) the temporal change of the direction
of the line-of-sight of the user and a temporal change of position
coordinates of a point on a screen of the display device closely
watched by the user, and (ii) the change of the position of the
presentation information on the screen of the display device.
11. A non-transitory recording medium storing a program causing a
computer to perform: acquiring a facial image sequence indicating a
user's face; displaying presentation information on a display
device and changing a position of the presentation information on a
screen of the display device; detecting, from the facial image
sequence, a direction of line-of-sight of the user; and determining
a face of the user indicated by the facial image sequence is a
living face, in accordance with relationship between temporal
change of the direction of the line-of-sight of the user and the
change of the position of the presentation information on the
screen of the display device.
12. The recording medium according to claim 11, wherein the
presentation information includes textual information.
13. The recording medium according to claim 11, wherein the
presentation information includes a diagram.
14. The recording medium according to claim 11, wherein the program
causes the computer to perform: determining a display timing and a
display position of the presentation information at random.
15. The recording medium according to claim 11, wherein the program
causes the computer to perform: determining the face of the user is
a living face, in accordance with relationship between (i) the
temporal change of the direction of the line-of-sight of the user
and a temporal change of position coordinates of a point on a
screen of the display device closely watched by the user, and (ii)
the change of the position of the presentation information on the
screen of the display device.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] The present application is a continuation application of
U.S. patent application Ser. No. 16/682,083 filed on Nov. 13, 2019,
which is continuation application of U.S. patent application Ser.
No. 15/514,573 filed on Mar. 27, 2017, which is issued as U.S. Pat.
No. 10,515,262, which is a National Stage Entry of international
application PCT/JP2015/005147, filed Oct. 9, 2015, which claims the
benefit of priority from Japanese Patent Application 2014-210529
filed on Oct. 15, 2014, the disclosures of all of which are
incorporated in their entirety by reference herein.
TECHNICAL FIELD
[0002] The present invention relates to a technology of detecting
an act of impersonation (spoofing) in face authentication.
BACKGROUND ART
[0003] A technology of detecting an act of spoofing in a security
system (face authentication system) using face authentication is
widely known. The act of spoofing is an act of pretending to be
another person to fraudulently utilize a system. The act of
spoofing is hereinafter also simply referred to as spoofing.
[0004] Spoofing against a face authentication system is performed
by an impersonator presenting another person's face different from
that of the impersonator to a camera for face imaging. The method
of presenting another person's face includes a method of presenting
a photograph of another person's face to a camera in the system.
Further, the method includes a method of presenting a video image
of another person's face displayed on a display of a smartphone or
the like to a camera in the system. Examples of related arts
relating to such spoofing are described in PTLs 1 to 8 below.
[0005] For example, a related art described in PTL 1 detects a
line-of-sight of a user from an image captured from a camera, while
presenting an image for drawing the user's attention on a display
device. Then, the related art (liveness detection) determines that
the user is a living body when the line-of-sight of the user points
in a direction of the display device, otherwise determines that
spoofing exists.
[0006] Further, a related art described in PTL 2 acquires a
plurality of facial images of a user respectively captured from a
plurality of different lines of sight. Then, the related art
extracts feature points in the user's face from the respective
plurality of facial images and determines whether or not the
feature points are on a same plane. Then, the related art
determines whether or not spoofing exists, in accordance with
whether or not the feature points are on a same plane.
[0007] Further, a related art described in PTL 3 measures distances
and angles to a plurality of measured points in a user's face by a
distance sensor to specify positions of the plurality of measured
points. Then, the related art determines whether or not spoofing
exists, in accordance with whether or not the plurality of measured
points are on a same plane.
[0008] Further, a related art described in PTL 4 guides a user to
move a line-of-sight by randomly moving a mark displayed on a
display device. Then, the related art determines whether or not
spoofing exists, in accordance with whether or not the
line-of-sight movement of the user agrees with a track of the
mark.
[0009] Further, a related art described in PTL 5 causes a user to
move his or her face in a direction instructed on a display screen
and captures the movement with a camera. Then, the related art
determines whether or not spoofing exists, in accordance with
whether or not a movement of a feature point can be detected in
time-series images capturing the user.
[0010] Further, a related art described in PTL 6 instructs a user
on a way to move a predetermined part of his or her face (e.g.
opening and closing of an eye) and captures the user's face. Then,
the related art determines whether or not spoofing exists, in
accordance with whether or not the predetermined part of the face
in an obtained dynamic image changes in accordance with the
instruction.
[0011] Further, a related art described in PTL 7 captures a
line-of-sight movement of a user moving in a presented image as a
dynamic line-of-sight pattern including time-series data of a
moving speed and a moved position. Then, the related art registers
the movement associated with a user identifier. Then, the related
art checks the user's dynamic line-of-sight pattern being an
authentication target against a registered dynamic line-of-sight
pattern to evaluate a degree of matching. By use of the related
art, whether or not spoofing exists can be determined in accordance
with the degree of matching of the dynamic line-of-sight
patterns.
[0012] Further, PTL 8 describes a line-of-sight detection
technology that can be employed in the respective related arts
described in PTLs 1, 4, and 7. The technology specifies a pupil
included in a facial image and detects a line-of-sight from a
position of the pupil and a direction of the face.
[0013] Thus, as a spoofing detection technology, a method of
determination by whether or not a line-of-sight points in a
direction of a display device is proposed. A method of
determination by whether or not a subject of imaging is a
three-dimensional object, and a method of determination by whether
or not a line-of-sight and a movement of a face, as instructed or
as registered, can be detected, are also proposed.
CITATION LIST
Patent Literature
[0014] [PTL1] Japanese Patent No. 4807167 [0015] [PTL2] Japanese
Patent No. 5035467 [0016] [PTL3] Japanese Patent No. 4734980 [0017]
[PTL4] Japanese Patent No. 4470663 [0018] [PTL5] Japanese
Unexamined Patent Application Publication No. 2008-305400 [0019]
[PTL6] Japanese Unexamined Patent Application Publication No.
2008-276345 [0020] [PTL7] Japanese Unexamined Patent Application
Publication No. 2007-141002 [0021] [PTL8] Japanese Unexamined
Patent Application Publication No. 2006-23953
SUMMARY OF INVENTION
Technical Problem
[0022] However, the related art described in PTL 1 is not able to
sufficiently detect spoofing caused by presenting a video image.
For example, it is assumed that an impersonator presents a video
image of another person's face captured so as to point in various
line-of-sight directions, being displayed on a display of a
smartphone or the like, to a camera in a face authentication
system. In this case, a likelihood that a line-of-sight direction
in the video image at some time point to a display device is high.
In this case, the related art mistakenly determines the face
displayed on the display as a living body. Further, when positions
of the display device and the camera in the face authentication
system are previously known, it is possible to prepare a video
image of a face, a line-of-sight of which points to the display
device when the image points to the camera. Thus, the related art
is not able to accurately detect spoofing caused by presenting a
video image.
[0023] Further, the related art described in PTL 2 requires a
plurality of cameras or a moving mechanism of a camera, in order to
acquire a plurality of facial images respectively captured from a
plurality of lines of sight. Further, the related art described in
PTL 3 requires a distance sensor in order to specify positions of a
plurality of measured points. Accordingly, there is a problem that
a scale of a device configuration increases in the related
arts.
[0024] Further, the related art described in PTL 4 guides a user to
track, by a line-of-sight, a random movement of a mark displayed
for spoofing determination. Further, the related art described in
PTL 5 instructs a user to move his or her face in a designated
direction for spoofing determination. Further, the related art
described in PTL 6 instructs a user to move a predetermined part of
his or her face for spoofing determination. Thus, a user is
required to perform an extra action for spoofing determination in
the related arts. Accordingly, the related arts increase a burden
on a user.
[0025] Further, the related art described in PTL 7 requires
preregistration of a dynamic line-of-sight pattern, and increases a
burden on a user.
[0026] Further, PTL 8 does not describe applying the line-of-sight
detection technology to spoofing determination.
[0027] The present invention is made to solve the aforementioned
problems. That is to say, an object of the present invention is to
provide a technology of more accurately detecting spoofing in face
authentication, without increasing a scale of a device
configuration and a burden on a user.
Solution to Problem
[0028] To achieve the above object, a spoofing detection device of
the present invention includes: facial image sequence acquisition
means that acquires a facial image sequence indicating a user's
face; line-of-sight change detection means that detects information
about temporal line-of-sight change from the facial image sequence;
presentation information display means that displays presentation
information presented to the user as part of authentication
processing; and spoofing determination means that determines a
likelihood that a face indicated by the facial image sequence is
spoofing, in accordance with the information about temporal
line-of-sight change with respect to the presentation
information.
[0029] A spoofing detection method of the present invention
includes: detecting, by a computer device, information about
temporal line-of-sight change from a facial image sequence
indicating a user's face; and determining, by a computer device, a
likelihood that a face indicated by the facial image sequence is
spoofing, in accordance with the information about temporal
line-of-sight change with respect to presentation information
presented to the user as part of authentication processing.
[0030] A recording medium, of the present invention, storing a
spoofing detection program causing a computer device to perform: a
facial image sequence acquisition step of acquiring a facial image
sequence indicating a user's face; a line-of-sight change detection
step of detecting information about temporal line-of-sight change
from the facial image sequence; an information presentation step of
displaying presentation information presented to the user as part
of authentication processing; and a spoofing determination step of
determining a likelihood that a face indicated by the facial image
sequence is spoofing, in accordance with the information about
temporal line-of-sight change with respect to the presentation
information.
Advantageous Effects of Invention
[0031] The present invention is able to provide a technology of
more accurately detecting spoofing in face authentication, without
increasing a scale of a device configuration and a burden on a
user.
BRIEF DESCRIPTION OF DRAWINGS
[0032] FIG. 1 is a functional block diagram of a spoofing detection
device according to a first example embodiment of the present
invention.
[0033] FIG. 2 is a diagram illustrating an example of a hardware
configuration of the spoofing detection device according to the
first example embodiment of the present invention.
[0034] FIG. 3 is a flowchart illustrating an operation of the
spoofing detection device according to the first example embodiment
of the present invention.
[0035] FIG. 4 is a functional block diagram of a spoofing detection
device according to a second example embodiment of the present
invention.
[0036] FIG. 5 is a flowchart illustrating an operation of the
spoofing detection device according to the second example
embodiment of the present invention.
[0037] FIG. 6 is a diagram illustrating an example of a facial
image sequence according to the second example embodiment of the
present invention.
[0038] FIG. 7 is a diagram illustrating an example of detected
change in a line-of-sight position, according to the second example
embodiment of the present invention.
[0039] FIG. 8 is a diagram illustrating an example of a detected
derivative value of a line-of-sight position, according to the
second example embodiment of the present invention.
[0040] FIG. 9 is a diagram illustrating an example of detected
information about temporal line-of-sight change, according to the
second example embodiment of the present invention.
[0041] FIG. 10 is a diagram illustrating an example of switching of
presentation information according to the second example embodiment
of the present invention.
[0042] FIG. 11 is a functional block diagram of a spoofing
detection device according to a third example embodiment of the
present invention.
[0043] FIG. 12 is a flowchart illustrating an operation of the
spoofing detection device according to the third example embodiment
of the present invention.
DESCRIPTION OF EMBODIMENTS
[0044] Example embodiments of the present invention will be
described in detail below with reference to the drawings.
First Example Embodiment
[0045] FIG. 1 illustrates a configuration of a spoofing detection
device 1 according to a first example embodiment of the present
invention. In FIG. 1, the spoofing detection device 1 includes a
facial image sequence acquisition unit 11, a line-of-sight change
detection unit 12, a presentation information display unit 13, and
a spoofing determination unit 14. In FIG. 1, a direction of an
arrow connecting rectangles representing functional blocks
indicates an example, and does not limit a signal direction between
functional blocks.
[0046] FIG. 2 illustrates an example of a hardware configuration of
the spoofing detection device 1. In FIG. 2, the spoofing detection
device 1 can be configured with a computer device. The computer
device includes a central processing unit (CPU) 1001, a random
access memory (RAM) 1002, a read only memory (ROM) 1003, a storage
device 1004, an imaging device 1005, and a display device 1006. The
ROM 1003 and the storage device 1004 store a computer program
causing the computer device to operate as the spoofing detection
device 1 according to the present example embodiment, and various
types of data. The imaging device 1005 is a device such as a camera
capturing an image of a subject and generating an image displaying
the subject. The display device 1006 is a device displaying various
types of information on a display screen. The CPU 1001 controls the
respective units in the computer device by reading the computer
program and the various types of data stored in the ROM 1003 and
the storage device 1004 into the RAM 1002, and executing the
program. In this case, the facial image sequence acquisition unit
11 is configured with the imaging device 1005, and the CPU 1001
reading the computer program and the various types of data stored
in the ROM 1003 and the storage device 1004 into the RAM 1002, and
executing the program. Further, the line-of-sight change detection
unit 12 is configured with the CPU 1001 reading the computer
program and the various types of data stored in the ROM 1003 and
the storage device 1004 into the RAM 1002, and executing the
program. Further, the presentation information display unit 13 is
configured with the display device 1006, and the CPU 1001 reading
the computer program and the various types of data stored in the
ROM 1003 and the storage device 1004 into the RAM 1002, and
executing the program. The spoofing determination unit 14 is
configured with the display device 1006, and the CPU 1001 reading
the computer program and the various types of data stored in the
ROM 1003 and the storage device 1004 into the RAM 1002, and
executing the program. The hardware configurations of the spoofing
detection device 1 and the respective functional blocks thereof are
not limited to the configuration described above.
[0047] The facial image sequence acquisition unit 11 acquires a
facial image sequence as input. The facial image sequence is a time
series of a plurality of facial images indicating a user's face.
The facial image sequence acquisition unit 11 may acquire the
facial image sequence through the imaging device 1005. Further, the
facial image sequence acquisition unit 11 may acquire, as the
facial image sequence, a facial image sequence in which a region
indicating a face is previously extracted. Alternatively, the
facial image sequence acquisition unit 11 may extract respective
facial regions from a time series of a plurality of images captured
by the imaging device 1005, and generate the facial image
sequence.
[0048] The line-of-sight change detection unit 12 detects
information about temporal line-of-sight change from a facial image
sequence. Specifically, the line-of-sight change detection unit 12
detects information indicating a line-of-sight from each facial
image constituting the facial image sequence. Then, the
line-of-sight change detection unit 12 may assume a time series of
information indicating a line-of-sight or a time derivative value
thereof as the information about temporal line-of-sight change.
Further, the line-of-sight change detection unit 12 may use both of
a time series of information indicating a line-of-sight and a
derivative value thereof as the information about temporal
line-of-sight change.
[0049] For example, the line-of-sight change detection unit 12 may
employ information indicating a line-of-sight direction as the
information indicating a line-of-sight. The information indicating
a line-of-sight direction may be information indicating a direction
of a line-of-sight on the basis of an optical axis of a camera, or
information indicating a direction in a global coordinate system.
Alternatively, the line-of-sight change detection unit 12 may
employ information indicating a line-of-sight position as the
information indicating a line-of-sight. For example, the
information indicating a line-of-sight position may be position
coordinates of a point closely watched by a user, in a display
screen in the display device 1006.
[0050] The line-of-sight change detection unit 12 may employ a
known line-of-sight detection technology for processing of
detecting information indicating a line-of-sight from a facial
image. When the employed line-of-sight detection technology assumes
a gray-scale image composed of brightness value information as a
facial image being a detection target, the aforementioned facial
image sequence acquisition unit 11 acquires a gray-scale facial
image sequence. Even in this case, an externally input image string
is not necessarily gray-scale. For example, an image string input
from the imaging device 1005 may be composed of RGB images composed
of color components of R (red), G (green), and B (blue). In this
case, the facial image sequence acquisition unit 11 or the
line-of-sight change detection unit 12 may perform processing of
generating a gray-scale image string from the input RGB image
string. Without being limited to the above, the facial image
sequence acquisition unit 11 or the line-of-sight change detection
unit 12 may acquire or generate a facial image sequence in a form
related to an employed line-of-sight detection technology.
[0051] The presentation information display unit 13 displays
presentation information on the display device 1006. The
presentation information refers to information presented to a user
as part of authentication processing. Further, the part of
authentication processing refers to presentation information being
presented to a user in a flow of authentication processing
performed in an unillustrated face authentication system in which
the spoofing detection device 1 is used. Such presentation
information is information presented as part of authentication
processing even when spoofing detection is not performed in such a
face authentication system. In other words, it is desirable that
such presentation information be not information separately
prepared for spoofing detection. For example, such presentation
information may include textual information simply indicating a
procedure such as "AUTHENTICATION STARTS," "PLEASE FACE FORWARD,"
and "PLEASE ENTER PASSCODE NUMBER." In addition, the presentation
information may include a diagram. The presentation information is
not limited to textual information and a diagram, and may include
another type of information, as long as the information is
presented to a user as part of authentication processing.
[0052] Specifically, the presentation information display unit 13
displays presentation information at any timing and at any display
position. The presentation information display unit 13 may
predetermine a display timing and a display position of the
presentation information. Alternatively, the presentation
information display unit 13 may randomly determine a display timing
and a display position for each processing operation performing
spoofing determination.
[0053] The spoofing determination unit 14 determines a likelihood
that a face indicated by a facial image sequence is spoofing, in
accordance with information about temporal line-of-sight change
with respect to presentation information. A line-of-sight of a user
changes in accordance with presentation information. For example,
the line-of-sight of the user is considered to change in accordance
with a display timing, a display position, a presentation content,
or the like of the presentation information. Further, for example,
the line-of-sight of the user is also considered to change in
accordance with a change in a display position of the presentation
information, a change in a presentation content, or the like.
Accordingly, the information about temporal line-of-sight change is
related to the presentation information. Accordingly, the spoofing
determination unit 14 may determine a likelihood of spoofing in
accordance with relevance of the information about temporal
line-of-sight change detected along with the presentation of the
presentation information to the presentation information.
[0054] Further, the spoofing determination unit 14 outputs a
determination result of spoofing. For example, when determining
that spoofing exists, the spoofing determination unit 14 may output
the determination result to the display device 1006 and another
output device (unillustrated).
[0055] An operation of the spoofing detection device 1 as
configured above will be described with reference to FIG. 3.
[0056] First, the facial image sequence acquisition unit 11 starts
acquisition of a facial image sequence (Step S1).
[0057] For example, the facial image sequence acquisition unit 11
acquires a facial image sequence by successively acquiring frames
constituting a video image input from the imaging device 1005 while
extracting a region indicating a face from each frame.
[0058] Next, the line-of-sight change detection unit 12 starts
detection of information about temporal line-of-sight change from
the facial image sequence (Step S2).
[0059] As described above, the line-of-sight change detection unit
12 detects information indicating a line-of-sight from each facial
image in the facial image sequence successively acquired by the
facial image sequence acquisition unit 11. Then, the line-of-sight
change detection unit 12 detects a time series and a derivative
value of the information as information about line-of-sight
change.
[0060] Next, the presentation information display unit 13 displays,
on the display device 1006, presentation information presented to a
user as part of authentication processing (Step S3).
[0061] Next, the spoofing determination unit 14 determines a
likelihood that a person indicated by the facial image sequence is
spoofing, in accordance with relevance of the information about
temporal line-of-sight change detected by the line-of-sight change
detection unit 12 to the presentation information displayed by the
presentation information display unit 13 (Step S4).
[0062] For example, the line-of-sight change detection unit 12 may
determine whether or not spoofing exists, in accordance with
whether or not the information about temporal line-of-sight change
relates to a presentation timing, a display position, or a content
of the presentation information, or changes thereof.
[0063] The above concludes the operation of the spoofing detection
device 1.
[0064] Next, an effect of the first example embodiment of the
present invention will be described.
[0065] The spoofing detection device according to the first example
embodiment of the present invention is able to more accurately
detect spoofing in face authentication, without increasing a scale
of a device configuration and a burden on a user.
[0066] The reason is that the facial image sequence acquisition
unit acquires a facial image sequence indicating a user's face, and
the line-of-sight change detection unit detects information about
temporal line-of-sight change from the facial image sequence.
Further, the presentation information display unit displays
presentation information presented to the user as part of
authentication processing. Then, the spoofing determination unit
determines a likelihood that the face included in the facial image
sequence is spoofing, in accordance with the information about
temporal line-of-sight change with respect to the presentation
information.
[0067] A line-of-sight of a real person not being spoofing changes
in accordance with presentation information. For example, a
line-of-sight of a real person changes at a timing when
presentation information is displayed on the display device and in
a direction of a display position in the display device. However,
when an impersonator displays a video image of a face of another
person different from himself or herself on a display of a
smartphone or the like, and presents the image to the imaging
device, a line-of-sight of the another person in the display
changes independently of the presentation information. For example,
a line-of-sight of the another person in the display changes
independently of a presentation timing and a display position of
the presentation information.
[0068] Specifically, when a line-of-sight of the another person in
the display is fixed, the line-of-sight does not change in
conjunction with a display timing of the presentation information.
Further, when a line-of-sight of the another person's face in the
display moves in various directions, the line-of-sight is highly
likely to change in a period in which the presentation information
is not presented or at a timing different from a presentation
timing. Further, in this case, the line-of-sight is highly likely
to point in a direction different from a direction of a display
position at a presentation timing. Consequently, when a video image
of another person's face is displayed on a display of a smartphone
or the like and the image is presented to a camera, information
about temporal line-of-sight change detected by the line-of-sight
change detection unit does not change in accordance with
presentation information. Accordingly, the present example
embodiment is able to accurately determine a likelihood of spoofing
in accordance with relevance of information about temporal
line-of-sight change to presentation information.
[0069] At this time, the present example embodiment uses
information about temporal change obtained from a natural movement
of a line-of-sight of a user to presentation information presented
as part of authentication processing. Accordingly, the user is not
required to perform an extra action or a line-of-sight movement,
being instructed for spoofing determination. Further, the present
example embodiment does not require a device, a distance sensor, or
the like for acquiring images from a plurality of lines of sight
for determination of three-dimensionality of a subject of imaging.
The present example embodiment may acquire a facial image sequence
of a user by at least one imaging device. Accordingly, the present
example embodiment does not increase a scale of a device
configuration.
Second Example Embodiment
[0070] Next, a second example embodiment of the present invention
will be described in detail with reference to drawings. In the
respective drawings referred to in the description of the present
example embodiment, a same component as and a step operating
similarly to the first example embodiment of the present invention
are respectively given same reference signs as the first example
embodiment, and detailed description thereof is omitted in the
present example embodiment.
[0071] First, FIG. 4 illustrates a configuration of a spoofing
detection device 2 according to the second example embodiment of
the present invention. In FIG. 4, the spoofing detection device 2
differs from the spoofing detection device 1 according to the first
example embodiment of the present invention in that a presentation
information display unit 23 is included in place of the
presentation information display unit 13. The spoofing detection
device 2 also differs from the spoofing detection device 1 in that
a spoofing determination unit 24 is included in place of the
spoofing determination unit 14. In FIG. 4, a direction of an arrow
connecting rectangles representing functional blocks indicates an
example, and does not limit a signal direction between functional
blocks.
[0072] The presentation information display unit 23 switches
presentation information presented to a user as part of
authentication processing. For example, the presentation
information display unit 23 may switch a display position of
presentation information. It is assumed that the switching of a
display position refers to changing a display position in a display
screen in a display device 1006. Further, for example, the
presentation information display unit 23 may switch presentation
information to another piece of presentation information. It is
assumed that the switching to another piece of presentation
information refers to changing presentation information displayed
on the display device 1006 to another piece of presentation
information. Further, when switching presentation information to
another piece of presentation information, the presentation
information display unit 23 may switch a display position thereof.
While the presentation information display unit 23 performs such
switching at least once, the switching frequency is not limited.
Further, the presentation information display unit 23 may perform
the switching of presentation information as described above at any
timing.
[0073] Specifically, for example, the presentation information
display unit 23 displays first presentation information at a timing
of a time m.sub.t1 at a display position (m.sub.x1,m.sub.y1). Then,
the presentation information display unit 23 switches the first
presentation information to second presentation information at a
timing of a time m.sub.t2. At this time, the presentation
information display unit 23 displays the second presentation
information at a display position (m.sub.x2,m.sub.y2). In this
case, the presentation information display unit 23 may predetermine
or randomly determine m.sub.t1, m.sub.t2, (m.sub.x1,m.sub.y1), and
(m.sub.x2,m.sub.y2).
[0074] The spoofing determination unit 24 determines a likelihood
of spoofing in accordance with relevance of information about
temporal line-of-sight change detected by the line-of-sight change
detection unit 12 to switching of presentation information.
Specifically, the spoofing determination unit 24 may determine
whether or not spoofing exists, by comparing the information about
temporal line-of-sight change detected from a facial image sequence
with a switching timing of the presentation information and a
change in a display position due to the switching. For example, the
spoofing determination unit 24 may determine whether or not a face
indicated by the facial image sequence is spoofing, in accordance
with whether or not the information about temporal line-of-sight
change relates to the aforementioned switching timing m.sub.t2.
Further, the spoofing determination unit 24 may perform spoofing
determination in accordance with whether or not the information
about temporal line-of-sight change relates to a change in the
display position by the switching from (m.sub.x1,m.sub.y1) to
(m.sub.x2,m.sub.y2).
[0075] An operation of the spoofing detection device 2 as
configured above will be described with reference to FIG. 5.
[0076] First, similarly to the first example embodiment of the
present invention, the spoofing detection device 2 performs Steps
S1 to S2. Thus, the spoofing detection device 2 starts acquisition
of a facial image sequence and starts detection of information
about temporal line-of-sight change.
[0077] Next, the presentation information display unit 23 displays,
on the display device 1006, first presentation information
presented to a user as part of authentication processing (Step
S23).
[0078] Next, the presentation information display unit 23 switches
the first presentation information to second presentation
information (Step S24).
[0079] Next, the spoofing determination unit 24 determines a
likelihood of spoofing with respect to a person indicated by the
facial image sequence, in accordance with relevance of the
information about temporal line-of-sight change detected by the
line-of-sight change detection unit 12 to the switching of the
presentation information by the presentation information display
unit 23 (Step S25).
[0080] For example, as described above, the line-of-sight change
detection unit 12 may determine whether or not spoofing exists, in
accordance with whether or not the information about temporal
line-of-sight change relates to a switching timing of the
presentation information and a change in a display position
accompanying the switching.
[0081] The above concludes the operation of the spoofing detection
device 2.
[0082] Next, the operation according to the second example
embodiment of the present invention will be described by use of a
specific example.
[0083] First, the facial image sequence acquisition unit 11 starts
acquisition of a facial image sequence (Step S1).
[0084] It is assumed that the facial image sequence acquired by the
facial image sequence acquisition unit 11 is a string of facial
images obtained by extracting face regions from respective
gray-scale images captured at certain intervals. For example, the
facial image sequence acquisition unit 11 may obtain a face
position by use of a known face detection technology from
time-series gray-scale images output from an imaging device 1005
such as a video camera. Then, the facial image sequence acquisition
unit 11 may extract face regions from the respective images as a
facial image sequence, in accordance with the obtained face
position. Alternatively, the facial image sequence acquisition unit
11 may acquire a facial image sequence obtained by previously
extracting face regions by an external device from an image string
output from the imaging device 1005. FIG. 6 illustrates an example
of a facial image sequence in this specific example. The facial
image sequence illustrated in FIG. 6 is acquired over a period from
a time 1 to a time T.
[0085] The line-of-sight change detection unit 12 detects
information about temporal line-of-sight change from the facial
image sequence in FIG. 6 (Step S2).
[0086] The line-of-sight change detection unit 12 detects
information about a line-of-sight by use of a known line-of-sight
detection technology from the respective facial images in the
facial image sequence. For example, when a known technology of
detecting a line-of-sight direction is employed, the line-of-sight
change detection unit 12 may detect a line-of-sight direction from
a facial image at a time t in the facial image sequence. Then, the
x component of the direction may be denoted as gx(t) and the y
component thereof as gy(t). Further, for example, when a known
technology of detecting a line-of-sight position is employed, the
line-of-sight change detection unit 12 may detect a line-of-sight
position from a facial image at a time tin the facial image
sequence. Then, the x coordinate of the position may be denoted as
gx(t) and the y coordinate thereof as gy(t). In this specific
example, it is assumed that information indicating a line-of-sight
position is employed.
[0087] Then, the line-of-sight change detection unit 12 uses time
series of gx(t) and gy(t) for T facial images from the start (t=1)
to the end (t=T) of the facial image sequence as information about
temporal line-of-sight change. Further, the line-of-sight change
detection unit 12 uses values [Gx(t),Gy(t)] obtained by
differentiating [gx(t),gy(t)] (t=1 to T) with respect to time t as
the information about temporal line-of-sight change. In other
words, in this specific example, the information about temporal
line-of-sight change is composed of four values with respect to
time t, a line-of-sight position [gx(t),gy(t)] and derivative
values thereof [Gx(t),Gy(t)].
[0088] For example, it is assumed that information illustrated in
FIGS. 7 to 9 is detected as information about temporal
line-of-sight change. FIG. 7 illustrates a line-of-sight position
[gx(t),gy(t)] detected from the respective facial images in the
facial image sequence. FIG. 8 illustrates values [Gx(t),Gy(t)]
obtained by differentiating the line-of-sight position
[gx(t),gy(t)] in FIG. 7 with respect to time t. FIG. 9 indicates
the line-of-sight position [gx(t),gy(t)] at each time t and the
derivative values [Gx(t),Gy(t)] thereof. As illustrated in FIGS. 7
to 9, the line-of-sight position detected from the facial image
sequence and the derivative values thereof change temporally.
[0089] Next, the presentation information display unit 23 displays
first presentation information at a timing of a time m.sub.t1 at a
display position (m.sub.x1,m.sub.y1) (Step S23). FIG. 10
illustrates a display example of the first presentation
information. In a diagram on the left in FIG. 10, the first
presentation information is textual information "AUTHENTICATION
STARTS." In this step, it is assumed that the presentation
information display unit 23 randomly selects the first presentation
timing m.sub.t1 and the display position (m.sub.x1,m.sub.y1).
[0090] Next, the presentation information display unit 23 switches
the first presentation information to second presentation
information at a timing of a time m.sub.t2 and displays the second
presentation information at a display position (m.sub.x2,m.sub.y2)
(Step S24). It is assumed that m.sub.t2>m.sub.t1. FIG. 10
illustrates a display example of the second presentation
information. In a diagram on the right in FIG. 10, the second
presentation information is textual information "PLEASE FACE
STRAIGHT TOWARD CAMERA." In this step, it is assumed that the
presentation information display unit 23 randomly selects the
second presentation timing m.sub.t2 and the display position
(m.sub.x2,m.sub.y2).
[0091] While two pieces of presentation information are switched
and displayed in this specific example, a case that three or more
pieces of presentation information are successively switched and
displayed is also applicable.
[0092] Next, the spoofing determination unit 24 performs spoofing
determination in accordance with relevance of the information about
temporal line-of-sight change detected by the line-of-sight change
detection unit 12 to the switching of the presentation information
by the presentation information display unit 23 (Step S25).
[0093] The spoofing determination unit 24 performs spoofing
determination by use of equations (1) to (3) below. As described
above, a line-of-sight position is denoted as [gx(t),gy(t)], and
derivative values thereof are denoted as [Gx(t),Gy(t)], as detected
information about temporal line-of-sight change. Further, it is
assumed that a time at which the first presentation information is
displayed is denoted as m.sub.t1 and a presentation position
thereof as (m.sub.x1, m.sub.y1). It is also assumed that a time at
which the second presentation information is displayed is denoted
as m.sub.t2 and a presentation position thereof as (m.sub.x2,
m.sub.y2).
[0094] First, the spoofing determination unit 24 calculates a time
.tau. at which a change in the line-of-sight is greatest around the
switching time m.sub.t2 of the presentation information as
expressed in equation (1) below.
.tau. = arg .times. .times. max m t .times. .times. 2 - t a
.ltoreq. t .ltoreq. m t .times. .times. 2 + t b .times. ( G x ( t )
+ .lamda. 1 .times. G y ( t ) ) ( 1 ) ##EQU00001##
[0095] G.sub.x.sup.(t) and G.sub.y.sup.(t) are the aforementioned
derivative values Gx(t) and Gy(t), respectively. Further, t.sub.a,
t.sub.b, and .lamda..sub.1 are parameters. Note that .lamda..sub.1
is a parameter for adjusting a balance between a change component
in an x direction and a change component in a y direction. Further,
t.sub.a and t.sub.b are parameters for determining an interval in
which .tau. is obtained.
[0096] Next, the spoofing determination unit 24 obtains
.epsilon..sub.1 being difference between magnitude of a change in
the line-of-sight position, and magnitude of a change in the
display position between the two pieces of presentation
information, on the basis of the time .tau. at which a change in
the line-of-sight is greatest, by use of equation (2) below.
Further, .epsilon..sub.1 is hereinafter referred to as an
evaluation value 1.
1 = abs .function. ( ( G x ( .tau. ) , G y ( .tau. ) ) T - .lamda.
2 .times. ( ( m x .times. .times. 2 , m y .times. .times. 2 ) T - (
m x .times. .times. 1 , m y .times. .times. 1 ) T ) ) ( 2 )
##EQU00002##
[0097] Note that .lamda..sub.2 is a parameter for adjusting a
balance between magnitude of a change in the line-of-sight
position, and magnitude of a change in the display position between
the two pieces of presentation information. Further, abs in
equation (2) is a function for calculating an absolute value.
[0098] Further, the spoofing determination unit 24 obtains
.epsilon..sub.2 being difference between a direction of a change in
the line-of-sight position, and a direction of a change in the
display position between the two pieces of presentation
information, by use of equation (3) below. Further, .epsilon..sub.2
is hereinafter referred to as an evaluation value 2.
2 = acos .function. ( ( G x ( .tau. ) , G y ( .tau. ) ) .times. ( (
m x .times. .times. 2 , m y .times. .times. 2 ) T - ( m x .times.
.times. 1 , m y .times. .times. 1 ) T ) ( G x ( .tau. ) , G y (
.tau. ) ) T - .lamda. 2 .times. ( ( m x .times. .times. 2 , m y
.times. .times. 2 ) T - ( m x .times. .times. 1 , m y .times.
.times. 1 ) T ) ) ( 3 ) ##EQU00003##
[0099] Then, the spoofing determination unit 24 compares the
evaluation value 1 with a threshold value 1 and compares the
evaluation value 2 with a threshold value 2. Then, the spoofing
determination unit 24 determines whether or not the face included
in the acquired facial image sequence is spoofing, in accordance
with the comparison results.
[0100] In the specific example described above, the values of the
evaluation values 1 and 2 respectively become smaller as the facial
image sequence acquired by the facial image sequence acquisition
unit 11 is more likely to indicate a real person's face. Further,
the values of the evaluation values 1 and 2 are respectively
designed to become larger as the facial image sequence is more
likely to indicate an impersonated face. The impersonated face
refers to a face or the like included in a video image presented on
a display of a smartphone or the like. Then, when both of the
evaluation values 1 and 2 are respectively greater than the
threshold values, the spoofing determination unit 24 determines
that the face is spoofing. Further, when at least either one of the
evaluation values 1 and 2 is less than the threshold value, the
spoofing determination unit 24 determines that the face is a
person's face (not spoofing). The spoofing determination unit 24
may determine that the face is spoofing when at least either one of
the evaluation values 1 and 2 is greater than the threshold
value.
[0101] Further, the threshold value 1 and the threshold value 2 may
be changed depending on a purpose. For example, for a purpose in
which failing to catch an impersonating person is fatal, it is
desirable that the values of the threshold value 1 and the
threshold value 2 be set on the low side. By contrast, for a
purpose in which a real person being mistakenly determined as
spoofing is a problem, it is desirable that the values of the
threshold value 1 and the threshold value 2 be set on the high
side.
[0102] Next, an effect of the second example embodiment of the
present invention will be described.
[0103] The spoofing detection device according to the second
example embodiment of the present invention is able to more
accurately detect spoofing in face authentication, without
increasing a scale of a device configuration and a burden on a
user.
[0104] The reason is that, in addition to a configuration similar
to the first example embodiment of the present invention, the
presentation information display unit switches and displays
presentation information presented to a user as part of
authentication processing. Then, the spoofing determination unit
determines whether or not a face included in a facial image
sequence is spoofing, in accordance with information about temporal
line-of-sight change with respect to the switching of the
presentation information.
[0105] A line-of-sight of a real person is considered to change
along with switching of presentation information in face
authentication. However, when an impersonator displays a video
image of another person's face different from himself or herself on
a display of a smartphone or the like and presents the image to the
imaging device, a line-of-sight of the another person's face in the
display changes independently of the switching of the presentation
information. Specifically, for example, when a line-of-sight of the
another person in the display is fixed, the line-of-sight does not
change at a switching timing of the presentation information.
Further, when a line-of-sight of the another person in the display
moves in a various manner, the line-of-sight is highly likely to
change in a period other than a switching timing of the
presentation information. Further, in this case, even when a
display position changes along with switching of the presentation
information, the line-of-sight is highly likely to change in a
direction of a position other than a display position after the
change. Accordingly, when a video image of another person's face is
displayed on a display of a smartphone or the like and the image is
presented to a camera, relevance of information about temporal
line-of-sight change detected by the line-of-sight change detection
unit to switching of presentation information is low.
[0106] Accordingly, the present example embodiment is able to
accurately determine spoofing in accordance with relevance of
information about temporal line-of-sight change to switching of
presentation information. Further, at this time, the present
example embodiment uses information about temporal line-of-sight
change with respect to switching of presentation information
presented as part of authentication processing. In general,
presentation information presented as part of authentication
processing often includes a plurality of types. Accordingly, by
utilizing information about temporal line-of-sight change with
respect to switching of such a plurality of pieces of presentation
information, the present example embodiment is able to apply a more
natural line-of-sight movement of a user in authentication
processing to spoofing determination. Consequently, the present
example embodiment reduces a burden on the user.
[0107] An example of the presentation information display unit
according to the present example embodiment switching presentation
information to another piece of presentation information has been
mainly described. In addition, the presentation information display
unit may perform switching of displaying same presentation
information at another display position at any timing.
Third Example Embodiment
[0108] Next, a third example embodiment of the present invention
will be described in detail with reference to drawings. In the
respective drawings referred to in the description of the present
example embodiment, a same component as and a step operating
similarly to the first example embodiment of the present invention
are respectively given same reference signs as the first example
embodiment, and detailed description thereof is omitted in the
present example embodiment.
[0109] First, FIG. 11 illustrates a configuration of a spoofing
detection device 3 according to the third example embodiment of the
present invention. In FIG. 11, the spoofing detection device 3
differs from the spoofing detection device 1 according to the first
example embodiment of the present invention in that a spoofing
determination unit 34 is included in place of the spoofing
determination unit 14. The spoofing detection device 3 also differs
from the spoofing detection device 1 in that a standard information
storage unit 35 is further included. The standard information
storage unit 35 constitutes an example embodiment of part of the
spoofing determination unit according to the present invention.
Further, in FIG. 11, a direction of an arrow connecting rectangles
representing functional blocks indicates an example, and does not
limit a signal direction between functional blocks.
[0110] The standard information storage unit 35 stores standard
information about temporal line-of-sight change related to a
content of presentation information. In general, a user is
considered to unconsciously or habitually move a line-of-sight in
accordance with a content of presentation information. The standard
information about temporal line-of-sight change indicates temporal
change in the line-of-sight that the person unconsciously or
habitually moves when the presentation information is
displayed.
[0111] For example, with respect to presentation information
composed of a character string "PLEASE ENTER PASSCODE NUMBER" and a
passcode number entry screen, a user is considered to move a
line-of-sight little by little near the passcode number entry
screen in order to enter a passcode number. Further, for example,
with respect to presentation information being a character string
"PLEASE FACE STRAIGHT TOWARD CAMERA," a user is considered to fix a
line-of-sight in a direction of a camera for a certain time, in
order to stare in the direction of the camera. Further, for
example, with respect to presentation information indicating a long
sentence, a user is considered to move a line-of-sight so as to
sequentially read words in the sentence, in order to read and
interpret the sentence. Thus, the standard information storage unit
35 stores generally assumed information about temporal
line-of-sight change, being associated with a content of the
presentation information.
[0112] The spoofing determination unit 34 compares information
about temporal line-of-sight change detected by the line-of-sight
change detection unit 12 with information about line-of-sight
change stored in the standard information storage unit 35, being
associated with a content of presentation information displayed at
the time. Then, when determining, in accordance with the comparison
result, that the detected information about temporal line-of-sight
change does not agree with the standard information about temporal
line-of-sight change related to the content of the presentation
information, the spoofing determination unit 34 determines that
spoofing exists. Further, when determining, in accordance with the
comparison result, that the two agree with one another, the
spoofing determination 34 determines that spoofing does not
exist.
[0113] An operation of the spoofing detection device 3 as
configured above will be described with reference to FIG. 12.
[0114] First, similarly to the first example embodiment of the
present invention, the spoofing detection device 3 performs Steps
S1 to S3. Thus, the spoofing detection device 3 starts acquisition
of a facial image sequence and detection of information about
temporal line-of-sight change, and displays presentation
information.
[0115] Next, the spoofing determination unit 34 compares the
information about temporal line-of-sight change detected by the
line-of-sight change detection unit 12 with standard information
about temporal line-of-sight change related to a content of the
presentation information presented by the presentation information
display unit 13. Then, in accordance with the comparison result,
the spoofing determination unit 34 determines a likelihood of
spoofing with respect to a person indicated by the facial image
sequence (Step S34).
[0116] The above concludes the operation of the spoofing detection
device 3.
[0117] Next, the operation according to the third example
embodiment of the present invention will be described by use of a
specific example.
[0118] In this specific example, the standard information storage
unit 35 prestores standard information about temporal line-of-sight
change, being associated with a content of assumed presentation
information. For example, the standard information storage unit 35
stores information about temporal change in a line-of-sight moving
little by little near a passcode number entry screen, being
associated with presentation information composed of textual
information "PLEASE ENTER PASSCODE NUMBER." Further, for example,
the standard information storage unit 35 stores information about
temporal change in a line-of-sight staring in a direction of a
camera for a certain time, being associated with presentation
information composed of textual information "PLEASE FACE STRAIGHT
TOWARD CAMERA." Further, for example, the standard information
storage unit 35 stores information about temporal change in a
line-of-sight sequentially moving through each word in a sentence
in a position of presentation information, being associated with
the presentation information composed of a long sentence. It is
assumed that, similarly to detected information about temporal
line-of-sight change, standard information about temporal
line-of-sight change is composed of a time series of a
line-of-sight position and derivative values thereof.
[0119] In this specific example, the spoofing detection device 3
operates similarly to the specific example according to the second
example embodiment of the present invention up to Steps S1 and S2.
Accordingly, it is assumed that the spoofing detection device 3
acquires a facial image sequence as illustrated in FIG. 6 and
detects information about temporal line-of-sight change as
illustrated in FIGS. 7 to 9.
[0120] Next, the presentation information display unit 13 displays
presentation information on a display device 1006 (Step S3).
[0121] Next, the spoofing determination unit 34 compares the
information about temporal line-of-sight change detected in Step S2
with standard information about temporal line-of-sight change
related to the presentation information displayed in Step S3. Then,
in accordance with the comparison result, the spoofing
determination unit 34 determines a likelihood of spoofing with
respect to a person indicated by the facial image sequence (Step
S34).
[0122] Specifically, the spoofing determination unit 34 calculates
difference .epsilon..sub.3 between a variance of the detected
information about temporal line-of-sight change [gx(t),gy(t)] and a
variance of the standard information about temporal line-of-sight
change [hx(t),hy(t)] by use of equation (4) below. Further,
.epsilon..sub.3 is hereinafter referred to as an evaluation value
3.
3 = var m t - t c .ltoreq. t .ltoreq. m t + t d .function. ( g x (
t ) ) - var m t - t c .ltoreq. t .ltoreq. m t + t d .function. ( h
x ( t ) ) + .lamda. 3 .times. var m t - t c .ltoreq. t .ltoreq. m t
+ t d .function. ( g y ( t ) ) - var m t - t c .ltoreq. t .ltoreq.
m t + t d .function. ( h y ( t ) ) ( 4 ) ##EQU00004##
[0123] Note that g.sub.x.sup.(t), g.sub.y.sup.(t), h.sub.x.sup.(t),
and h.sub.y.sup.(t) are the aforementioned information about
temporal change gx(t), gy(t), hx(t), and hy(t), respectively.
Further, var is a function for calculating a variance. Further,
.lamda..sub.3 is a parameter for adjusting a balance between an x
component and a y component. Further, m.sub.t denotes a time at
which the presentation information is displayed by the presentation
information display unit 13. Further, t.sub.c and t.sub.d are
parameters for determining an interval in which a variance is
calculated. Each of the values of t.sub.c and t.sub.d may be set to
a positive value, a negative value, or zero. For example, when both
values of t.sub.c and t.sub.d are set to positive values, equation
(4) above calculates the evaluation value 3 with respect to a
predetermined interval including the time m.sub.t. Further, when
t.sub.c is set to zero and t.sub.d is set to a positive value,
equation (4) above calculates the evaluation value 3 with respect
to a predetermined interval at and after the time m.sub.t. Further,
when t.sub.c is set to a negative value and t.sub.d is set to a
positive value, equation (4) above calculates variances with
respect to a predetermined interval at or after the time m.sub.t
considering a human reaction rate.
[0124] The spoofing determination unit 34 compares the evaluation
value 3 with a threshold value 3. Then, when the evaluation value 3
is greater than the threshold value 3, the spoofing determination
unit 34 determines that a face indicated by the facial image
sequence is spoofing. Further, when the evaluation value 3 is less
than the threshold value 3, the spoofing determination unit 34
determines that the face indicated by the facial image sequence is
not spoofing.
[0125] Further, the present example embodiment may be implemented
in combination with the second example embodiment of the present
invention. In that case, the spoofing detection device 3 includes
the presentation information display unit 23 according to the
second example embodiment of the present invention in place of the
presentation information display unit 13. Then, similarly to the
spoofing determination unit 24 according to the second example
embodiment of the present invention, the spoofing determination
unit 34 may calculate the evaluation values 1 and 2 in the specific
example described above. Additionally, the spoofing determination
unit 34 may calculate the evaluation value 3 in the specific
example described above. In this case, the spoofing determination
unit 34 may calculate the evaluation value 3 by comparing detected
information about temporal line-of-sight change with standard
information about temporal line-of-sight change with respect to
presentation information before and after switching, respectively.
Then, in accordance with the comparison result between the
evaluation value 1, the evaluation value 2, and the evaluation
value 3, and the respective threshold values, the spoofing
determination unit 34 may perform spoofing determination. For
example, the spoofing determination unit 34 may determine that the
face is spoofing when all of the evaluation value 1, the evaluation
value 2, and the evaluation value 3 are greater than the respective
threshold values.
[0126] Next, an effect of the third example embodiment of the
present invention will be described.
[0127] The spoofing detection device according to the third example
embodiment of the present invention is able to more accurately
detect spoofing in face authentication, without increasing a scale
of a device configuration and a burden on a user.
[0128] The reason is that, in addition to a configuration similar
to the first example embodiment of the present invention, the
spoofing determination unit determines a likelihood of spoofing
with respect to a face indicated by a facial image sequence, in
accordance with relevance of detected information about temporal
line-of-sight change to a content of presentation information.
[0129] In face authentication, a line-of-sight of a real person is
considered to change unconsciously in accordance with a content of
presentation information. However, when an impersonator displays a
video image of another person's face different from himself or
herself on a display of a smartphone or the like and presents the
image to an imaging device, a line-of-sight of the another person's
face in the display is highly likely to exhibit a movement not
assumed from the content of the presentation information.
Specifically, for example, in a case that a line-of-sight of the
another person in the display is fixed, even when the content of
the presentation information is a content prompting entry of a
passcode number, the line-of-sight of the another person in the
display does not change. Further, in a case that a line-of-sight of
the another person in the display moves in a various manner, even
when the content of the presentation information is, for example, a
sentence, a change in the line-of-sight of the another person in
the display is highly likely to be different from a movement of a
line-of-sight reading words in the sentence. Accordingly, when a
video image of another person's face is displayed on a display of a
smartphone or the like and the image is presented to a camera,
information about temporal line-of-sight change detected by the
line-of-sight change detection unit does not agree with standard
information about temporal line-of-sight change related to a
content of presentation information.
[0130] Thus, the present example embodiment is able to determine
spoofing by use of a natural movement of a line-of-sight of a user,
the movement being related to a content of presentation information
presented to the user as part of authentication processing.
Consequently, the present example embodiment is able to accurately
detect spoofing, without increasing a burden on the user for
spoofing detection.
[0131] The facial image sequence acquisition unit according to the
respective aforementioned example embodiments of the present
invention has been described on the assumption that a facial image
sequence based on images input from the imaging device is acquired
nearly in real time. In addition, the facial image sequence
acquisition unit may acquire a facial image sequence stored in a
storage device (e.g. video data captured and saved by an imaging
device in the past). In this case, the spoofing determination unit
may acquire information about a presentation timing and a
presentation position of presentation information displayed by the
presentation information display unit at the time of capturing the
facial image sequence. For example, information about such
presentation information may be stored in the storage device along
with the facial image sequence. Then, the spoofing determination
unit may compare information indicating temporal line-of-sight
change detected from the facial image sequence acquired from the
storage device with information about the presentation
information.
[0132] Further, a case of using a time series of information
indicating a line-of-sight position or a line-of-sight direction,
and derivative values thereof as information about temporal
line-of-sight change according to the respective aforementioned
example embodiments of the present invention has been mainly
described. In addition, the information about temporal
line-of-sight change may be another type of information as long as
the information can be detected from a facial image sequence and
can indicate temporal change in a line-of-sight.
[0133] Further, an example of the spoofing determination unit
according to the aforementioned second and third example
embodiments of the present invention determining spoofing by use of
the evaluation values 1 to 3 has been mainly described. At this
time, equations for respectively calculating the evaluation values
1 to 3 are not limited to the above. Further, without being limited
to the evaluation values, the spoofing determination unit may
determine spoofing in accordance with relevance of detected
information about temporal line-of-sight change to presentation
information presented to a user as part of authentication
processing.
[0134] Further, an example of the spoofing determination unit
according to the respective aforementioned example embodiments of
the present invention determining whether or not a face indicated
by a facial image sequence is spoofing has been mainly described.
In addition, the spoofing determination unit may output information
indicating a degree of spoofing as a determination result.
[0135] Further, an example of each functional block in the spoofing
detection device according to the respective aforementioned example
embodiments of the present invention being provided by a CPU
executing a computer program stored in the storage device or a ROM
has been mainly described. In addition, the respective functional
blocks may be provided, in part, in whole, or in combination, by
dedicated hardware.
[0136] Further, the functional blocks of the spoofing detection
device according to the respective aforementioned example
embodiments of the present invention may be provided by a plurality
of devices in a distributed manner.
[0137] Further, the operations of the spoofing detection device
according to the respective aforementioned example embodiments of
the present invention described with reference to the flowcharts
may be stored in the storage device (recording medium) in a
computer device as a computer program according to the present
invention. Then, the CPU may read out and execute such a computer
program. Then, in such a case, the present invention is composed of
a code of such a computer program or a recording medium.
[0138] Further, the respective aforementioned example embodiments
may be implemented in combination as appropriate.
[0139] Further, without being limited to the respective
aforementioned example embodiments, the present invention may be
implemented in various embodiments.
INDUSTRIAL APPLICABILITY
[0140] The spoofing detection technology according to the present
invention is able to more accurately detect spoofing in face
authentication, without increasing a scale of a device
configuration and a burden on a user. The technology is widely
applicable to a purpose of improving a security level of a system
using face authentication.
[0141] The present invention has been described with the
aforementioned example embodiments as exemplary examples. However,
the present invention is not limited to the aforementioned example
embodiments. In other words, various embodiments that can be
understood by a person skilled in the art may be applied to the
present invention, within the scope thereof.
[0142] This application claims priority based on Japanese Patent
Application No. 2014-210529 filed on Oct. 15, 2014, the disclosure
of which is hereby incorporated by reference thereto in its
entirety.
REFERENCE SIGNS LIST
[0143] 1, 2, 3 Spoofing detection device [0144] 11 Facial image
sequence acquisition unit [0145] 12 Line of sight change detection
unit [0146] 13, 23 Presentation information display unit [0147] 14,
24, 34 Spoofing determination unit [0148] 35 Standard information
storage unit [0149] 1001 CPU [0150] 1002 RAM [0151] 1003 ROM [0152]
1004 Storage device [0153] 1005 Imaging device [0154] 1006 Display
device
* * * * *